Fitting to predict anatomy

Hello,

Currently, I am making a SSM of the humerus. The next step is to fit the SSM on a model to predict/fit the complete anatomy. I wonder if this is easy to do in Shapeworks. Is it necessary to use the DeepSSM function for this? Or can the SSM be exported to fit the model on different softwares (e.g. 3-Matic)? How many coding skills do you need for this?

Extra information about my research: I would like to determine the original/complete anatomy of humeri that have a fracture. So I will fit the part of the humerus that is still intact on the SSM to determine the original anatomy of the other half of the humerus.

Does someone have experience?
Thank you in advance.
Cheers,
Florianne

Hello
I’m also working on this, I’m curious to see if I can export a statistical shape model of variability through shapeworks, this model contains the information needed for deformation, and then I write my own program to change it so that this model fits my structure.
If you have done your job with Shapeworks, please let me know that it is able to accomplish this task, it is very important for me, thank you.

The shape model is entirely contained in the saved particle files. Each particle is in correspondence across shapes.

ShapeWorks Studio will read these particle files and compute the PCA to perform dimension reduction and perform other analyses, but these particle files could be read in other software to perform other statistical analysis.

Surface reconstruction is performed in Studio by using a template mesh reference (median shape per the model) and warping it using the correspondence points via LibIGL.

Determining the original/complete anatomy from a fracture is likely possible, but we don’t have a pre-defined workflow for this task.

Hello,

I use shapeworks to create SSM, can this transformation information be stored in a .h5 file or other types of files, because I don’t want to operate in Studio, I want to use C++ or Python programs to operate on this .h5 file, change the parameters of different principal components, and then generate new models. Because I want to collect some discrete points on the tibia surface and use these sparse points to predict the shape of the model. Can this be achieved with shapeworks?
Thank you very much for your answer!

The resulting particle files are very simply X Y Z, one particle per line. The inputs are segmentations (any 3d volume type ITK reads) or meshes (any mesh format VTK reads). The projects can be stored as json files. The grooming and optimizing can be done via Python or CLI.

Hello, I used four STL type files to generate a statistical shape model, but the following problem arises, I want to know how to avoid this problem? Thank you for your answer.

This appears to be a poor shape model with swapped correspondences. I would start with a much lower count (e.g. 32 particles) and increase the number of iterations per split. After verifying that the particles are in correspondence across shapes, start increasing the particle count.

Thank you very much.I will try .